Predicting and Explaining Drug Toxicity with Graph Neural Networks (available)

This student research project focuses on predicting and explaining drug toxicity using advanced machine learning techniques, specifically Graph Neural Networks (GNNs).

Graph Neural Networks are a modern, powerful class of machine learning models widely used across diverse fields, from chemistry to social networks. However, one of the major challenges with traditional GNNs is their “black-box” nature, where the functions they implement are complex and difficult to interpret. In this project, the student will explore how monotonic GNNs can bridge this gap by providing more understandable explanations for their predictions, which is essential in critical fields like drug toxicity prediction. This research could contribute to making GNNs more accessible and trustworthy for real-world applications, especially in fields where interpretability and regulatory compliance are paramount.

The student will implement a monotonic GNN, a specialised type of GNN, to analyse a real-world database of drug candidates and predict their toxicity levels. Monotonic GNNs are unique in that they offer a crucial advantage: their predictions can be fully explained using logical, interpretable rules, making them more transparent than traditional GNNs. The student will extract these rules from the GNN model and compare them with the pre-existing explanations in the dataset, providing insights into the interpretability of machine learning models in the context of drug development.